20:00:18 #startmeeting glance 20:00:19 Meeting started Thu Jul 10 20:00:18 2014 UTC and is due to finish in 60 minutes. The chair is markwash. Information about MeetBot at http://wiki.debian.org/MeetBot. 20:00:20 Useful Commands: #action #agreed #help #info #idea #link #topic #startvote. 20:00:22 The meeting name has been set to 'glance' 20:00:24 o/ 20:00:33 o/ 20:00:33 o/ 20:00:42 o/ 20:00:45 o/ 20:01:15 okay let's get started 20:01:24 agenda here: https://etherpad.openstack.org/p/glance-team-meeting-agenda 20:01:29 #topic bug day 20:01:35 we got a little heat on the mailing list recently 20:01:55 jokke_: since you added the item to the agenda, here's the mic 20:01:57 * markwash passes the mic 20:02:28 oh 20:02:34 I guess he didn't o/ 20:02:35 haha 20:02:44 lol 20:02:52 ok 20:02:54 arnaud: any thoughts about bug day from your quarter? 20:02:55 I can take the mic 20:02:57 o/ 20:03:06 so based on https://etherpad.openstack.org/p/glance-bug-day 20:03:16 the bug day will happen Week 14-18 20:03:23 which is next week 20:03:31 I will send an email to the ML later today 20:03:39 great! 20:03:40 hello 20:03:40 sorry 20:03:48 something to add jokke_ ? 20:03:48 jokke_: no worries 20:03:52 for the bug day? 20:03:54 I have a business travel next week, sorry 20:04:07 no ... I was just hoping we get bit more attendance for that than you and me :) 20:04:12 :) 20:04:26 the ML post might recruit a few more 20:04:31 yep 20:04:33 what day in that week? 20:04:46 Mon or Tue or Wed 20:04:47 what's the purpose of bug day: a bug triage or bug fixing? 20:05:02 (or adding new bugs? :)) 20:05:06 triage mostly 20:05:31 what about Tuesday? 20:05:50 Tuesday is good for me 20:05:52 I'd vote for Wed ... might have some actions needed at the start of the week due change landing prod 20:05:53 Monday is a bit hard 20:06:02 Wednesday also works for me 20:06:08 ok so Wed 20:06:11 +1 -> Tuesday 20:06:13 :) 20:06:16 if that's ok with you guys 20:06:17 lol 20:06:18 bug fixing imo. ativelkov 20:06:29 I probably can do Tue as well 20:06:31 i can try to join at Wed 20:06:36 Mon I really rather not 20:06:42 I mean it can be Tuesday and Wednesday 20:06:49 +1 20:06:49 but you don't have to be 100% both days 20:06:51 +3 20:07:05 ok so let's do that then 20:07:14 I think we have enough for 2 days anyways 20:07:16 What are the hours? I'll be happy to join any day, but my time zone is quite far from yours 20:07:21 if not, happy days 20:07:38 ativelkov: which timezone you're at? 20:07:45 UTC+4 20:08:13 hi all 20:08:19 let's at least have a clear kickoff 20:08:22 ativelkov: we can start on your afternoon and I carry that over for the US folks when they wake up :P 20:08:26 so I know when to first show up 20:08:47 mine is utc+8 :P 20:08:56 jokke_: great, this will do 20:08:58 I'm currently on utc+1 20:09:21 So, we'll have 24 hours of bugs :) 20:09:26 utc-7 20:09:40 relay 20:09:44 :) 20:09:55 markwash: you need to move 1000miles East :D 20:09:56 * nikhil___ remembers olympics 20:10:08 jokke_: more like 6K 20:11:16 why 1000 east? why not 14000 miles west? 20:11:23 okay 20:11:24 we have 265 open bugs which of 105 are still at New 20:11:48 so I'm pretty sure we have enough to do even passing the stick on 20:11:55 +1 20:12:30 also I think we need to do a better job at tagging 20:12:34 the bugs 20:12:42 ##openstack-glance-bugs? 20:12:56 9 propose-close 20:12:56 9 low-hanging-fruit 20:12:56 5 db 20:12:56 4 testing 20:12:59 ... 20:13:19 arnaud: we also would need to document the tags 20:13:38 good point 20:13:49 I'm guessing we'll get a lot out of propose-close as we go through that list 20:13:50 https://wiki.openstack.org/wiki/Bug_Tags has only propose-close after I put it there 20:13:59 markwash: hopefully 20:14:23 so, we're trying for two days? did I hear that correctly? 20:14:27 why no on #openstack-glance ? 20:15:08 yep markwash 20:15:11 nikhil___: I was just thinking to have the bug off from our normal daily chat + off from review announcements 20:15:12 one more question are those 2 days as per UTC time? 20:15:24 markwash: correct Tue+Wed 20:15:45 markwash: When do you want to do Kickoff? 20:15:56 jokke_: ah k 20:15:57 okay, sounds good, let see that email to the list, I think folks can respond with their general plans for the 2 day timeframe 20:16:25 #link https://wiki.openstack.org/wiki/Bug_Tags 20:16:37 jokke_: I'm not sure, I think we just want to keep people mutually engaged, so probably we can do that if people give a sense of when they'll start in a response on the ML 20:16:48 so let's pass on a kickoff 20:16:58 and just try to keep visibility high 20:17:09 sound okay? 20:17:12 #link https://etherpad.openstack.org/p/glance-bug-day 20:17:20 markwash: sounds good to me 20:17:40 okay great 20:17:43 next up! 20:17:55 #topic functional testing to tempest 20:18:13 jokke_: this is your item 20:18:22 but nikhil___ has some info too, I imagine 20:18:23 #link https://review.openstack.org/#/c/106146 20:18:41 I think nikhil___ you were driving this ... I just brought it to agenda 20:18:46 gotcha 20:18:47 thanks 20:18:59 anybody not on board with this plan? 20:19:47 jokke_: oh glad you brought it up 20:19:53 My plan was to ask around when we will start, but nikhil___ proposed the spec in between ... so my only quiestion would be timing? 20:20:10 think many people have interest in this 20:20:20 zhiyan: hemanth__ ^ ? 20:20:40 Will this move only glance/tests/functional? Or glance/tests/integration as well? 20:20:45 jokke_: oops 0:-) 20:20:52 nikhil___: yes 20:21:10 I think it will save the need for some of the legacy tests in glance/tests/functional, perhaps 20:21:10 nikhil___: sure 20:21:26 since the main goal of those legacy tests was retiring a lot of glance/tests/functional 20:21:52 but I strongly encourage people to second guess me on that position :-) 20:22:03 jokke_: zhiyan hemanth__ : haven't added any other assignees to that yet, if you all wanted to add your name and push a PS ? 20:22:34 I would prefer us not to throw all the functionals to tempest at least to start with 20:22:48 markwash: agree, tentatively that seems like the approach we want to take. 20:22:51 +1 :) 20:23:06 I think that's something we would need to discuss what funtionals and perhaps integrations we want to move 20:23:09 jokke_: +1 , if we can do it in staggered fashion 20:23:18 okay, so maybe we can take cleaning up integration tests as a followup? 20:23:25 +1 20:23:43 fine by me 20:24:13 back to the question of when 20:24:20 what's our goal? j2? 20:24:29 can we embed also re-enablement of some tests ... I've noticed that we have quite a few tests skipped with comments "skipped until bug 100xxxxx gets resolved" 20:24:30 j3? 20:24:30 Launchpad bug 100 in launchpad "uploading po file overwrites authors list" [Medium,Fix released] https://launchpad.net/bugs/100 20:24:42 +1 -> j3 20:24:47 pipe down uvirtbot 20:25:02 markwash: how about j2 non-voting j3 fully in game tests removed from codebase? 20:25:14 nikhil___: may i know where you want to move those test cases to tempest? api or scenario? 20:25:33 jokke_: only concern, I'd is not to raise TC's hope of adding more tests to tempest 20:25:33 or do we have ready place for them? 20:25:58 on the other hand whatever helps, I guess 20:25:58 I think the TC is only tracking the binary image data coverage part 20:26:18 they probably are happy leaving the rest of it to us 20:26:20 markwash: that's the one that is there already? :P 20:26:48 jokke_: sorry, what do you mean exactly? 20:26:54 markwash: while talking on the qe channel, I heard murmurs of not enough Glance tests in tempest and that someone got correlated with the wiki 20:27:27 jokke_: fwiw I like that timetable you proposed 20:27:33 s/someone/somehow/gc 20:27:38 I'd like to see some concrete progress in j2 20:27:46 markwash: I think it was last or previous week mentioned that the tempest tests for glance are already using binary data, no? 20:28:14 ah, perhaps 20:28:47 so I know nikhil___ is looking into that binary data testing stuff, he can confirm when he has enough info 20:28:59 ah 20:29:01 it sounded like there was at least something we should add to the tests to satisfy the gaps 20:29:08 (I just disconnected from irc and now back, by this id, hum, networking is kill me, again) 20:29:19 what markwash said :) 20:29:43 apart from that effort, who is looking to help with adding the glance functional tests to tempest? 20:29:53 that probably needs to happen before we remove the tests from glance 20:30:05 nikhil___: i just said i like join this efforts 20:30:12 zhiyan1: great 20:30:14 I need some pointers where to start with, but will be more than happy to help 20:30:24 okay I guess we'll see people patch their names into the glance-spec 20:30:31 so we'll have people and a timetable 20:30:33 sounds good to me 20:30:40 ready for next topic? 20:30:43 may i know where you want to move those test cases to tempest? api or scenario? 20:30:48 nikhil___: ^ 20:31:03 (i'm not sure i posted this msg successfully) 20:31:09 zhiyan1: I'm a little in the dark about tempest's structure--do you have a recommendation? 20:31:27 same here 20:31:28 frankly, me too. will check it 20:31:50 zhiyan1: markwash jokke_ : think we should plan to attend the openstack-qe team meeting later today to figure it out 20:31:58 or some/one of us :) 20:32:06 @ which time is that? 20:32:08 it's at 22UTC iirc 20:32:22 nikhil___: good idea 20:32:34 I'll see if I stay awake :D 20:32:38 I cannot make that unfortunately, but I look forward to hearing back :-) 20:32:52 if somebody wants to dive in and propose a change to tempest, though 20:32:58 I'm sure the tempest core folks would give us some pointers 20:33:07 if we accidentally picked the wrong spot the first time :-) 20:33:12 :P 20:33:14 markwash: np, last time mtreinish asked me to join to discuss a bit more however I was waiting on creating that spec to give a better picture :) 20:33:21 markwash guys you should make more rally benchmarks as well! 20:33:22 =) 20:33:49 boris-42: you're more than welcome to contribute ;D 20:33:55 jokke_ where?) 20:34:10 ^^ 20:34:11 * nikhil___ hides 20:34:18 lol 20:34:25 boris-42 has provided us the framework :-) 20:34:36 markwash and integration in gates 20:34:40 with reports=) 20:34:50 markwash: and apparently just volunteered to extend that :) 20:34:53 yeah, seems it's useful to catch perf issue 20:35:04 all right, let see if the details of tempest testing can coagulate around that spec, I think we can move on 20:35:08 markwash btw we have base support of SLA=) 20:35:12 boris-42, where can we find the reports? 20:35:22 arnaud https://github.com/stackforge/rally 20:35:26 #topic rally and osprofiler 20:35:33 boris-42: ping 20:35:33 :-) 20:35:36 markwash pong lol 20:35:37 =) 20:35:44 markwash so as I don't sleep=) 20:36:01 boris-42: can you link us in again to your glance osprofiler integration change 20:36:04 probably some of us know that I am annoying everybody with requests to merge my patches ASAP 20:36:05 =) 20:36:06 sorry, I lost the link 20:36:21 markwash https://github.com/stackforge/osprofiler 20:36:30 https://review.openstack.org/#/c/105635/ 20:36:31 ^ So this is the repo of project 20:36:52 and this is already merged in glance https://review.openstack.org/#/c/103362/ 20:37:06 So lemme introduce you guys in what is osprofilre 20:37:12 and what are future plans 20:37:16 zhiyan1: thanks! 20:37:48 Major difference from cProfiler and osprofiler that our goal is to build one trace of points 20:37:57 that goes through all services and projects 20:38:14 e.g. nova during booting of VM requests images from glance 20:38:30 okay it's simpler to show lol 20:38:41 +1 20:38:43 shoot 20:38:47 where we find an example of report? 20:38:52 the trace view is great 20:38:55 https://gist.github.com/boris-42/c3c3ee1c2c7db40de236 20:38:59 ^ this is plain text 20:39:08 e.g. json with tree of objects 20:39:29 in info goes everything from tracepoint that you would like to put 20:39:42 and this can be presented in different ways 20:39:51 boris-42: do you have a link to share that shows the overall timing view of a request? 20:39:52 http://pavlovic.me/rally/profiler/ one is this 20:40:03 markwash ^ you mean this ? 20:40:12 yes that's it 20:40:22 actually I am planing to use a bit different thing 20:40:30 http://bl.ocks.org/kerryrodden/7090426 20:40:37 ^ this seems much better 20:40:50 f a n c y 20:40:59 haha yes 20:41:06 so the major super cool stuff about profiler is 20:41:09 that it works out of box 20:41:17 it can be turned on always even in production 20:41:22 boris-42: that looks nifty 20:41:23 and it won't produce any load 20:41:32 unit it is triggered 20:41:42 by special headers in HTTP request 20:41:58 boris-42: the result looks pretty good , btw, that's the unit of "finished" and "start" field? 20:42:13 zhiyan1 milisecs 20:42:24 10^-3 secs 20:42:40 boris-42: thanks 20:42:53 and the key thing is those headers are actually secure in the sense that 20:43:01 markwash yep they are secure 20:43:12 markwash you have to know HMAC key that is specified in api-paste.ini 20:43:18 to be able to send proper headers 20:43:21 you have to have a private key in order to trigger a trace generation 20:43:25 yep yep 20:43:46 so with this private key situation, the idea is to have the default for osprofiler integration to be "on" 20:44:03 and that's the main thing I wanted people to know about 20:44:13 markwash and one more thing 20:44:15 boris-42: so the value of "SECRET_KEY" in glance-api-paste.ini you proposed is a hardcode value? or just a env name 20:44:16 if it is on by default 20:44:27 zhiyan1 it's hardcoded 20:44:36 zhiyan1 and should be the same in all projects 20:44:38 operator needs to change it as needed right? 20:44:42 ok 20:44:44 got 20:44:45 zhiyan1 yep like passwords 20:44:45 -) 20:44:53 don't use 12345 as a pass 20:44:54 =) 20:44:55 so is this trace done only on the requests where the trigger is? 20:45:00 jokke_ yep 20:45:08 Not like you put it on and need to remember to turn it off as well 20:45:09 jokke_ but it can be extended in future 20:45:13 oh cool 20:45:16 no I like that 20:45:19 ;) 20:45:23 I mean we can extend it 20:45:27 to trace every X request 20:45:35 and collect useful info for opertors 20:45:35 boris-42: so those trace data will be send out from glance by message right? 20:45:43 zhiyan1 it's first implementation 20:45:51 based on Ceilometer + oslo.messaging 20:46:07 in future probably we will be able to use something else 20:46:08 so it will spend some overhead , iiuc right? 20:46:22 zhiyan1 yep if it is triggered 20:46:27 zhiyan1 it will produce overhead 20:46:35 (lets give this a few more minutes, we need to spend some time on the Graffiti and Artifacts stuff) 20:46:49 markwash ok 20:46:52 I think if folks can follow up with questions on the review it would be most helpful 20:46:53 ok markwash 20:46:57 Btw 20:46:59 one more thing 20:47:12 as it stands, we have a config option deployers can use to turn *off* profiling completely 20:47:15 if it will be turned on by default we will be able to add support profiling in rally gates 20:47:26 and they can also (with more difficulty) rip it out of the paste config 20:47:50 so you will be able to get traces during benchmark in nice report 20:47:52 automatically 20:47:54 markwash: so what about disable it by default? (seems you like enable it) 20:48:05 e.g. put any amount of traces + put proper benchmark (via plugging) + and run it 20:48:11 and get all that you need=) 20:48:29 so we will get rid of holly wars about perfromance=) 20:48:37 I want it on by default so that its on in the gates and is generally useable by folks without having to restart services 20:48:58 and I think the crypto story is good enough to allow us to turn it on by default 20:48:59 zhiyan1 yep as it doesn't produce any overhhead 20:49:04 zhiyan1 why not keeping it on?) 20:49:27 okay, but I think I need to cut it off here, so we can have some time to check in with ativelkov and TravT 20:49:40 markwash ok good 20:49:47 thanks guys, let's follow up on the review 20:49:59 #topic Graffiti / Catalog 20:50:11 re: Graffiti 20:50:13 TravT, lakshmiS, et al: still around? 20:50:15 We've started posting code and I'm about to post an update to the spec based on review comments and things we've discovered in development. 20:50:26 Glance: #link: https://review.openstack.org/#/c/105904 20:50:32 python-glanceClient - #link: https://review.openstack.org/#/c/105231/ 20:50:49 And a coupld of Horizon ones in progress 20:51:07 #link: https://review.openstack.org/#/c/99761/ #link: https://review.openstack.org/#/c/104063/ 20:51:29 great TravT 20:51:53 Everything is work in progress, but we're always open to feedback. 20:52:03 excellent 20:52:10 I look forward to the update to the spec 20:52:39 one question that has come up is, we've added CRUD methods to glance/db/sqlalchemy/api.py...do we need to add similar support to the registry/api.py? 20:53:11 I think that is necessary, yes 20:53:12 wayne__: yes please 20:53:19 yes 20:53:21 ok 20:53:30 it would be nice to add them to the simple db api as well 20:53:31 and simple 20:53:41 and is the glance/db/simple used primarily for testing? 20:53:41 yes 20:54:12 any other questions or notes for now? 20:54:20 Some early reviews on the code submitted will be helpful since we are trying to get ourselves familiar 20:54:32 good point 20:54:47 okay, let's move along, only a little time left 20:54:52 #topic artifact repository 20:55:19 btw the mission statement change was approved 20:55:22 https://review.openstack.org/#/c/98002/ 20:55:27 ativelkov: any notes for us today? 20:55:34 Work is in progress here. Modeling the common metadata fileds. DB, migrations, API, all the stuff 20:55:35 \o/ 20:55:50 Will post first changesets soon and update the spec accordingly 20:55:57 * jokke_ appalauds 20:56:14 Got the idea about the in-plugin reference management 20:56:48 For now I want only the static references to exist, however the plugins will get the interface to do the queries during artifact imports 20:56:54 ativelkov: do you think that approach will work out, at least initially? 20:57:06 ativelkov: sound great, looking forward to view them 20:57:37 ativelkov: okay, so for now you're saying, artifacts can resolve dynamic lookups to static during import 20:57:49 markwash: exatcly 20:57:56 okay 20:58:11 Hope it should be fine for Heat 20:58:22 I might need to negotiate some breathing room around that, but I don't think you're moving in the wrong direction at all 20:58:33 so I think we're in good shape 20:58:45 s/need to/try to/ 20:58:46 :-) 20:58:46 If not, we'll have to provide a way for plugins to expose custom API methods 20:59:03 Which is doable but rises securty questions 20:59:24 * markwash nods 20:59:30 okay, thanks folks! 20:59:31 -1 on that if possible 20:59:35 #topic open discussion 20:59:40 1 minute! 20:59:46 actually we don't usually have followers in this late timeslot 20:59:47 thanks markwash that was productive hour 20:59:52 just a bunch of people who need to go to bed :-) 21:00:02 markwash: i need some edu, to know how to enable auto-config-gen in gate 21:00:35 let's move to glance irc room 21:00:41 Did we get any volunteer to fix https://bugs.launchpad.net/glance/+bug/1315321 ? I remember it was discussed last week but didn't see the result 21:00:42 Launchpad bug 1315321 in glance "image_size_cap not checked in v2" [Undecided,In progress] 21:01:05 ativelkov: hmm, not yet 21:01:34 thanks everybody 21:01:37 #endmeeting